A feature vector calculation method and related device based on artificial intelligence

Abstract

The embodiment of the present application discloses a method for calculating a feature vector. When calculating the feature vector of media data, a training data set is determined from the historical media data based on historical user behavior data, and the historical user behavior data includes historical media data that the user has clicked on. Based on the training data set, a feature vector of the historical media data is obtained using a vector prediction model. Since the type identifier can reflect the content characteristics of the historical media data, the type identifiers included in the training data set are statistically analyzed, and the feature vector of the type identifier is determined based on the feature vector of the historical media data corresponding to the type identifier. The feature vector of the type identifier can reflect the content characteristics of the historical media data. If target media data for which the feature vector is not calculated is obtained, and the target media data is similar to the historical media data with the same type identifier, the feature vector of the target media data can be calculated based on the feature vector of the type identifier possessed by the target media data.

Description

Technical Field

[0001] The present application relates to the field of computers, and in particular to an artificial intelligence-based media data recommendation method and related devices. Background Art

[0002] With the development of information technology, users can quickly and easily access a large amount of information such as media data. In order to provide users with the media data they need or are interested in, such as news, advertisements, videos, etc., a system for recommending media data to users has emerged on Internet websites.

[0003] Push based on media data similarity is the core of the recommendation system, and media data can be represented by the feature vector of media data. Therefore, in order to realize the recommendation of media data, it is usually necessary to calculate the feature vector of media data so as to determine the similarity of media data according to the similarity of feature vectors and realize the recommendation of media data.

[0004] Currently, feature vectors of media data are obtained through unsupervised training, and then the feature vectors of media data are used for recommendation tasks. However, when new media data is generated in the push system, this method cannot calculate the feature vectors of the newly generated media data, making it difficult to accurately implement recommendation services. Summary of the invention

[0005] In order to solve the above technical problems, the present application provides a feature vector calculation method and related devices based on artificial intelligence, which can calculate the feature vector of the target media data. Therefore, the feature vector user recommendation model of all media data can be trained to accurately implement the recommendation service, and the feature vector of the target media data can be calculated without relearning all the media data, avoiding the consumption of a lot of time and space.

[0006] The embodiments of the present application disclose the following technical solutions:

[0007] In a first aspect, an embodiment of the present application provides a method for calculating a feature vector based on artificial intelligence, the method comprising:

[0008] Determining a training data set from historical media data according to historical user behavior data, wherein the historical user behavior data includes historical media data clicked by a user;

[0009] According to the training data set, a feature vector of the historical media data is obtained using a vector prediction model;

[0010] Counting the type identifiers included in the training data set, where the type identifiers are used to reflect the content characteristics of the historical media data;

[0011] Determining a feature vector of the type identifier according to a feature vector of historical media data corresponding to the type identifier;

[0012] If target media data for which a feature vector has not been calculated is obtained, a feature vector of the target media data is calculated according to a feature vector of a type identifier of the target media data.

[0013] In a second aspect, an embodiment of the present application provides a feature vector calculation device, the device comprising a first determination unit, a second determination unit, a statistical unit, a third determination unit and a calculation unit:

[0014] The first determining unit is used to determine a training data set from historical media data according to historical user behavior data, wherein the historical user behavior data includes historical media data clicked by a user;

[0015] The second determining unit is configured to obtain a feature vector of the historical media data using a vector prediction model according to the training data set;

[0016] The statistical unit is used to count the type identifiers included in the training data set, and the type identifiers are used to reflect the content characteristics of the historical media data;

[0017] The third determining unit is used to determine the feature vector of the type identifier according to the feature vector of the historical media data corresponding to the type identifier;

[0018] The calculation unit is configured to calculate a feature vector of the target media data according to a feature vector of a type identifier of the target media data if target media data for which a feature vector has not been calculated is acquired.

[0019] In a third aspect, an embodiment of the present application provides a device for calculating a feature vector, the device comprising a processor and a memory:

[0020] The memory is used to store program code and transmit the program code to the processor;

[0021] The processor is configured to execute the method described in the first aspect according to instructions in the program code.

[0022] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium is used to store program code, and the program code is used to execute the method described in the first aspect.

[0023] It can be seen from the above technical solution that when calculating the feature vector of media data, a training data set is determined from the historical media data based on the historical user behavior data, and the historical user behavior data includes the historical media data that the user has clicked on. Based on the training data set, the feature vector of the historical media data is obtained using a vector prediction model. Since the type identifier can reflect the content characteristics of the historical media data, the type identifiers included in the training data set are statistically analyzed, and the feature vector of the type identifier is determined based on the feature vector of the historical media data corresponding to the type identifier. The feature vector of the type identifier can reflect the content characteristics of the historical media data. If the target media data for which the feature vector is not calculated is obtained, and the target media data is similar to the historical media data with the same type identifier, the feature vector of the target media data can be calculated based on the feature vector of the type identifier possessed by the target media data. Therefore, the training of the user recommendation model for the feature vectors of all media data can accurately implement the recommendation service, and at the same time, the feature vector of the target media data can be calculated without re-learning all the media data, avoiding the consumption of a lot of time and space. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0025] Figure 1 A schematic diagram of an application scenario of a feature vector calculation method provided in an embodiment of the present application;

[0026] Figure 2 A flow chart of a method for calculating a feature vector provided in an embodiment of the present application;

[0027] Figure 3 An example diagram of a video clicked by a user provided in an embodiment of the present application;

[0028] Figure 4 A flow chart of a method for calculating a feature vector provided in an embodiment of the present application;

[0029] Figure 5a A structural diagram of a feature vector calculation device provided in an embodiment of the present application;

[0030] Figure 5b A structural diagram of a feature vector calculation device provided in an embodiment of the present application;

[0031] Figure 5c A structural diagram of a feature vector calculation device provided in an embodiment of the present application;

[0032] Figure 6 A structural diagram of a feature vector calculation device provided in an embodiment of the present application;

[0033] Figure 7 A structural diagram of a server provided in an embodiment of the present application. DETAILED DESCRIPTION

[0034] The embodiments of the present application are described below in conjunction with the accompanying drawings.

[0035] In the recommendation system, a lot of new media data is added every day. Traditional methods obtain the feature vector of media data through unsupervised training on offline data sets, and then use the feature vector of media data for recommendation tasks. When new media data is added, it is impossible to calculate the feature vector of the new media data, making it difficult to accurately implement push services.

[0036] In order to solve the above technical problems, the embodiment of the present application provides a feature vector calculation method based on artificial intelligence. The method can calculate the feature vector of the target media data using the feature vector of the type identification of the target media data for a target media data for which the feature vector has not been calculated. In this way, the feature vector user recommendation model of all media data is trained so that the recommendation service can be accurately implemented. At the same time, the feature vector of the target media data can be calculated without relearning all the media data, avoiding the consumption of a lot of time and space.

[0037] It should be emphasized that the feature vector calculation method provided in the embodiment of the present application is based on artificial intelligence. Artificial Intelligence (AI) is a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines so that the machines have the functions of perception, reasoning and decision-making.

[0038] Artificial intelligence technology is a comprehensive discipline that covers a wide range of fields, including both hardware-level and software-level technologies. The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics and other technologies. Artificial intelligence software technology mainly includes computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0039] In the embodiments of the present application, the artificial intelligence software technologies mainly involved include the above-mentioned natural language processing and machine learning.

[0040] For example, it may involve natural language processing (NLP), text preprocessing, semantic understanding, and search-related aspects such as sorting, keywords, and recommendations.

[0041] For example, it may involve machine learning (ML), which is a multi-disciplinary interdisciplinary subject involving probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and other disciplines. It specializes in studying how computers simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications are spread across all areas of artificial intelligence. Machine learning usually includes technologies such as deep learning. Deep learning includes artificial neural networks, such as convolutional neural networks (CNN), recurrent neural networks (RNN), deep neural networks (DNN), etc.

[0042] The method can be applied to a data processing device, which can be a terminal device. The terminal device can be, for example, a smart terminal, a computer, a personal digital assistant (PDA), a tablet computer, or the like.

[0043] The data processing device may also be a server, and the server may be an independent server or a cluster server.

[0044] The feature vector calculation method provided in the embodiment of the present application can be applied to a scenario of media data recommendation, in which the feature vector of the media data calculated by a data processing device is used to recommend the media data to the user.

[0045] It can be understood that, on the product side, this method can be used in various media data recommendation software, and the software can be installed on a terminal device, so as to recommend media data to users based on the feature vectors of the media data calculated by this method.

[0046] In order to facilitate understanding of the technical solution of the present application, the feature vector calculation method provided in the embodiment of the present application is introduced below in combination with actual application scenarios, taking the data processing device as a server as an example.

[0047] See also Figure 1 , Figure 1 Schematic diagram of an application scenario of the feature vector calculation method provided in an embodiment of the present application. The application scenario includes a server 101 and a terminal device 102, and the server 101 recommends media data to the user based on the calculated feature vector through the terminal device 102. The media data can be in different forms, such as pictures, videos, news, advertisements, etc.

[0048] When calculating the feature vector of the media data, the server 101 may obtain historical user behavior data, where the historical user behavior data includes historical media data clicked by the user.

[0049] Server 101 determines a training data set from historical media data based on historical user behavior data, and obtains a feature vector of historical media data using a vector prediction model based on the training data set. Since the type identifier can reflect the content characteristics of the historical media data, server 101 can count the type identifiers included in the training data set, and determine the feature vector of the type identifier based on the feature vector of the historical media data corresponding to the type identifier. The feature vector of the type identifier can reflect the content characteristics of the historical media data. If server 101 obtains target media data for which a feature vector has not been calculated, and the target media data is similar to historical media data with the same type identifier, server 101 can calculate the feature vector of the target media data based on the feature vector of the type identifier possessed by the target media data. Thus, the feature vector user recommendation model of all media data is trained so that the recommendation service can be accurately implemented.

[0050] In this way, when the user corresponding to the terminal device 102 clicks on a certain media data, the server 101 can use the trained recommendation model to recommend to the user media data that other users may like based on the media data. Figure 1 The introduction is mainly based on various media data such as historical media data and target media data is video. Of course, the embodiment of the present application does not limit the form of the media data.

[0051] Next, the feature vector calculation method provided in the embodiment of the present application will be introduced in detail with reference to the accompanying drawings, taking the data processing device being a server as an example.

[0052] See also Figure 2 , Figure 2 A flow chart of a method for calculating a feature vector is shown, the method comprising:

[0053] S201. Determine a training data set from historical media data according to historical user behavior data.

[0054] The historical user behavior data may be the user's operation behavior on the historical media data, such as the behavior of clicking on the historical media data, wherein the historical user behavior data includes the historical media data clicked by the user.

[0055] In this embodiment, a possible way to determine the training data set is that the server can obtain historical user behavior data in batches from the product line, collect historical media data clicked by users, and combine all the historical media data clicked by a user into sentences of this user. Each unit in the sentence is a historical media data, and all the sentences of all users are combined into a training data set.

[0056] Taking the historical media data as an example, when users watch videos through video recommendation software, they can click on the videos they like. Figure 3 This is an example of some videos that users have clicked. The server can obtain all videos that users have clicked. The videos that a user has clicked constitute the user's sentences, and all the user's sentences constitute the training data set.

[0057] It is understandable that the user's click behavior on historical media data expresses the user's feelings before viewing the historical media data. For example, after seeing the title, cover and other information of the video (historical media data), the user wants to watch the video. However, after clicking on a video, the user may watch it for a while and find that it is not a video he likes, so he quickly swipes away. Or after watching a video, a user thinks the video is good and posts a comment, such as clicking a like.

[0058] The user's comment information reflects the user's degree of liking for historical media data, and the feature vector of historical media data is used to recommend media data to the user. Therefore, in order to make the feature vector of historical media data more accurately reflect the user's degree of liking for historical media data so as to recommend the user's favorite media data to the user, in a possible implementation method, if the historical media data has been viewed, the historical user behavior data also includes comment information.

[0059] In this case, the method of determining the training data set in S201 can be to determine the historical media data with positive comment information based on the comment information, thereby determining the training data set based on the historical media data with positive comment information. For example, in the training data set, the sampling rate of the historical media data with positive comment information is increased, so that the feature vector of the historical media data determined based on the training data set can accurately reflect the user's liking for the historical media data, thereby increasing the probability of recommending the media data that the user likes.

[0060] S202: Obtain a feature vector of the historical media data using a vector prediction model according to the training data set.

[0061] It should be noted that there is a certain contextual relationship between the historical media data clicked by the user. Therefore, in this embodiment, the vector prediction model can be a skip-gram model, that is, the feature vector (ie, embedding) of each historical media data is obtained by training using the skip-gram method.

[0062] S203: Count the type identifiers included in the training data set.

[0063] The server may count the type identifiers included in the training data set. The type identifier is used to reflect the content characteristics of the historical media data. The category identifier may be pre-set, and the category identifier may be, for example, a category or a label.

[0064] Category can identify the classification of historical media data. Category can be divided into primary category, secondary category, etc. Primary category can include games, life, technology, education, humanities, landscape, etc. Secondary category can be a subcategory under primary category, such as Honor of Kings, PlayerUnknown's Battlegrounds, etc. For example, if the historical media data is a video, and the video is a live video of the game "Honor of Kings", then the primary category of the video can be games, and the secondary category can be Honor of Kings.

[0065] The tag can identify the corresponding content of the historical media data. For example, the tags of the historical media data are stars and athletes. There can be one or more tags.

[0066] The training data set includes historical media data. Each historical media data may have a corresponding type identifier. The historical media data may have one or more of the above type identifiers. Therefore, the type identifiers included in the training data set may be counted.

[0067] For example, the training data set includes video 1 and video 2, where video 1 has a corresponding first-level category "life", and video 2 has a corresponding first-level category "game" and a second-level category "Honor of Kings". It can be determined that the training data set includes the first-level category "life", the first-level category "game" and the second-level category "Honor of Kings".

[0068] S204: Determine a feature vector of the type identifier according to the feature vector of the historical media data corresponding to the type identifier.

[0069] The server obtains the historical media data corresponding to each type identifier, and thereby determines the feature vector of the type identifier according to the feature vector of the historical media data corresponding to the type identifier.

[0070] In one case, the type identifier includes multiple historical media data. In this case, the server can obtain all historical media data corresponding to the type identifier from the training data set, and take the average of the feature vectors of all historical media data as the feature vector of the type identifier.

[0071] For example, the type identifier is a first-level category, and the historical media data is a video. For the category "life" under the first-level category, the average value of the feature vectors of all videos under the category "life" in the training data set is used as the feature vector of the category "life".

[0072] S205: If target media data whose feature vector has not been calculated is obtained, calculate a feature vector of the target media data according to a feature vector of a type identifier of the target media data.

[0073] When the server obtains the target media data, the server may obtain the type identifier of the target media data, and thereby calculate the feature vector of the target media data according to the feature vector of the type identifier, wherein the target media data is the media data for which the feature vector has not been calculated.

[0074] In some cases, the target media data may include multiple type identifiers. In this case, the server may calculate an average value of the feature vectors of the multiple type identifiers as the feature vector of the target media data.

[0075] For example, the target media data is a video, the first-level category of the target media data is "game", and the second-level category is "Honor of Kings", then the feature vectors of the first-level category and the second-level category can be averaged as the feature vector of the video.

[0076] It should be noted that in the embodiment of the present application, the target media data can be newly generated media data, or it can be media data whose feature vectors have not been calculated in historical media data. If the target media data is media data whose feature vectors have not been calculated in historical media data, then the target media data has corresponding historical user behavior data. At this time, after the server obtains the target media data, the feature vector of the target media data can be obtained by an incremental learning method based on the historical user behavior data of the target media data. That is, the feature vector of the target media data is obtained using incremental skip-gram.

[0077] It is understandable that, due to the addition of target media data, the degree to which each media data is liked by the user may change over time, and then its corresponding feature vector should also change to accurately reflect the user's liking of the media data. Therefore, in a possible implementation, if new user behavior data is generated for the target media data after a preset time, the feature vector of all media data can also be updated through an incremental learning method based on the new user behavior data. Among them, all media data includes the target media data and the historical media data.

[0078] After the feature vector is calculated, the feature vectors of all media data can be used to train a recommendation model. The recommendation model can predict which media data may be liked by the user based on the feature vectors of the media data, thereby recommending the media data that the user likes. In the process of recommendation, the recall of media data and / or the sorting of media data can be achieved.

[0079] The feature vector of the target media data obtained by the incremental learning method and the feature vector of the target media data obtained by using the feature vector of the type identifier can complement each other and be used together for the training and prediction of the recommendation model.

[0080] It can be seen from the above technical solution that when calculating the feature vector of media data, a training data set is determined from the historical media data based on the historical user behavior data, and the historical user behavior data includes the historical media data that the user has clicked on. Based on the training data set, the feature vector of the historical media data is obtained using the vector prediction model. If the target media data for which the feature vector is not calculated is obtained, since the type identifier can reflect the content characteristics of the historical media data, the type identifier included in the training data set is statistically analyzed, and the feature vector of the type identifier is determined based on the feature vector of the historical media data corresponding to the type identifier. The feature vector of the type identifier can reflect the content characteristics of the historical media data. If the target media data is similar to the historical media data with the same type identifier, the feature vector of the target media data can be calculated based on the feature vector of the type identifier of the target media data. Therefore, the training of the user recommendation model of the feature vectors of all media data can accurately realize the recommendation service, and at the same time, the feature vector of the target media data can be calculated without re-learning all the media data, avoiding the consumption of a lot of time and space.

[0081] In addition, the method provided in the embodiment of the present application effectively improves the offline model evaluation index (area under the curve, AUC) and improves online indicators such as click-through rate, penetration rate, and new user acquisition rate.

[0082] Next, the feature vector calculation method provided by the embodiment of the present application will be introduced in combination with an actual application scenario. In this application scenario, the media data is a video, and the obtained feature vector can be used for video recall and sorting. Figure 4 , the method comprising:

[0083] S401. The server collects videos clicked by users.

[0084] S402: The server composes the user's sentences from the videos clicked by the user to form a training data set.

[0085] S403: The server obtains a feature vector of the video through a skip-gram model according to the data training set.

[0086] S404: The server converts the feature vector of the video into feature vectors of categories and tags.

[0087] S405: The server collects data in the recommendation system.

[0088] S406: The server calculates a feature vector of the new video using the feature vectors of the categories and tags.

[0089] S407: The server calculates a feature vector of the new video using an incremental skip-gram model.

[0090] S408: The server uses the feature vector to perform recall and sorting.

[0091] Based on the feature vector calculation method provided in the above embodiment, the present application embodiment also provides a feature vector calculation device, see Figure 5a , the device includes a first determining unit 501, a second determining unit 502, a statistical unit 503, a third determining unit 504 and a calculating unit 505:

[0092] The first determining unit 501 is used to determine a training data set from historical media data according to historical user behavior data, wherein the historical user behavior data includes historical media data clicked by a user;

[0093] The second determining unit 502 is configured to obtain a feature vector of the historical media data using a vector prediction model according to the training data set;

[0094] The statistical unit 503 is used to count the type identifiers included in the training data set, where the type identifiers are used to reflect the content characteristics of the historical media data;

[0095] The third determining unit 504 is configured to determine a feature vector of the type identifier according to a feature vector of the historical media data corresponding to the type identifier;

[0096] The calculation unit 505 is configured to calculate a feature vector of the target media data according to a feature vector of a type identifier of the target media data if target media data for which a feature vector has not been calculated is acquired.

[0097] In a possible implementation, the target media data has historical user behavior data, see Figure 5b , the apparatus further includes a fourth determining unit 506:

[0098] The fourth determining unit 506 is configured to obtain a feature vector of the target media data by an incremental learning method according to the historical user behavior data of the target media data.

[0099] In a possible implementation, if new user behavior data is generated for the target media data after a preset time, see Figure 5c , the device further comprises an updating unit 507:

[0100] The updating unit 507 is used to update the feature vectors of all media data by an incremental learning method according to the new user behavior data, where the all media data includes the target media data and the historical media data.

[0101] In a possible implementation, if the historical media data corresponding to the type identifier includes multiple pieces, the third determining unit 504 is specifically configured to:

[0102] Acquire all historical media data corresponding to the type identifier from the training data set;

[0103] The feature vectors of all historical media data are averaged as the feature vector of the type identifier.

[0104] In a possible implementation, if the historical media data has been viewed, the historical user behavior data also includes comment information.

[0105] In a possible implementation manner, the first determining unit 501 is configured to:

[0106] Determining historical media data having positive comment information according to the comment information;

[0107] The training data set is determined according to historical media data having positive comment information.

[0108] In a possible implementation manner, the obtained feature vectors of all media data are used to train a recommendation model, where the all media data include the target media data and the historical media data.

[0109] The present application also provides a device for calculating feature vectors. The following describes the device for calculating feature vectors in conjunction with the accompanying drawings. Figure 6 As shown, the embodiment of the present application provides a device 600 for feature vector calculation, and the device 600 may also be a terminal device, and the terminal device may be any intelligent terminal including a mobile phone, a tablet computer, a personal digital assistant (PDA), a point of sales (POS), a vehicle-mounted computer, etc., and the terminal device is a mobile phone as an example:

[0110] Figure 6 FIG. 1 is a block diagram showing a partial structure of a mobile phone related to a terminal device provided in an embodiment of the present application. Figure 6 The mobile phone includes: a radio frequency (RF) circuit 610, a memory 620, an input unit 630, a display unit 640, a sensor 650, an audio circuit 660, a wireless fidelity (WiFi) module 670, a processor 680, and a power supply 690. Those skilled in the art will understand that Figure 6 The mobile phone structure shown in the figure does not constitute a limitation on the mobile phone, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0111] Combine the following Figure 6 A detailed introduction to the various components of the mobile phone:

[0112] The RF circuit 610 can be used for receiving and sending signals during the process of sending and receiving information or making calls. In particular, after receiving the downlink information of the base station, it is sent to the processor 680 for processing; in addition, the designed uplink data is sent to the base station. Usually, the RF circuit 610 includes but is not limited to an antenna, at least one amplifier, a transceiver, a coupler, a low noise amplifier (Low Noise Amplifier, referred to as LNA), a duplexer, etc. In addition, the RF circuit 610 can also communicate with the network and other devices through wireless communication. The above-mentioned wireless communication can use any communication standard or protocol, including but not limited to the Global System of Mobile communication (Global System of Mobile communication, referred to as GSM), General Packet Radio Service (General Packet Radio Service, referred to as GPRS), Code Division Multiple Access (Code Division Multiple Access, referred to as CDMA), Wideband Code Division Multiple Access (Wideband Code Division Multiple Access, referred to as WCDMA), Long Term Evolution (Long Term Evolution, referred to as LTE), email, Short Messaging Service (Short Messaging Service, referred to as SMS), etc.

[0113] The memory 620 can be used to store software programs and modules. The processor 680 executes various functional applications and data processing of the mobile phone by running the software programs and modules stored in the memory 620. The memory 620 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory 620 can include a high-speed random access memory, and can also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0114] The input unit 630 can be used to receive input digital or character information, and to generate key signal input related to the user settings and function control of the mobile phone. Specifically, the input unit 630 may include a touch panel 631 and other input devices 632. The touch panel 631, also known as a touch screen, can collect the user's touch operation on or near it (such as the user's operation on the touch panel 631 or near the touch panel 631 using any suitable object or accessory such as a finger, stylus, etc.), and drive the corresponding connection device according to a pre-set program. Optionally, the touch panel 631 may include two parts: a touch detection device and a touch controller. Among them, the touch detection device detects the user's touch orientation, detects the signal brought by the touch operation, and transmits the signal to the touch controller; the touch controller receives the touch information from the touch detection device, converts it into contact coordinates, and then sends it to the processor 680, and can receive and execute commands sent by the processor 680. In addition, the touch panel 631 can be implemented in various types such as resistive, capacitive, infrared, and surface acoustic waves. In addition to the touch panel 631, the input unit 630 may also include other input devices 632. Specifically, the other input devices 632 may include but are not limited to one or more of a physical keyboard, function keys (such as volume control keys, switch keys, etc.), a trackball, a mouse, a joystick, etc.

[0115] The display unit 640 can be used to display information input by the user or information provided to the user and various menus of the mobile phone. The display unit 640 may include a display panel 641. Optionally, the display panel 641 may be configured in the form of a liquid crystal display (LCD), an organic light-emitting diode (OLED), etc. Further, the touch panel 631 may cover the display panel 641. When the touch panel 631 detects a touch operation on or near it, it is transmitted to the processor 680 to determine the type of touch event. Subsequently, the processor 680 provides a corresponding visual output on the display panel 641 according to the type of touch event. Although in Figure 6 In the embodiment, the touch panel 631 and the display panel 641 are used as two independent components to realize the input and output functions of the mobile phone, but in some embodiments, the touch panel 631 and the display panel 641 can be integrated to realize the input and output functions of the mobile phone.

[0116] The mobile phone may also include at least one sensor 650, such as a light sensor, a motion sensor, and other sensors. Specifically, the light sensor may include an ambient light sensor and a proximity sensor, wherein the ambient light sensor may adjust the brightness of the display panel 641 according to the brightness of the ambient light, and the proximity sensor may turn off the display panel 641 and / or the backlight when the mobile phone is moved to the ear. As a type of motion sensor, the accelerometer sensor can detect the magnitude of acceleration in all directions (generally three axes), and can detect the magnitude and direction of gravity when stationary. It can be used for applications that identify the posture of the mobile phone (such as horizontal and vertical screen switching, related games, magnetometer posture calibration), vibration recognition related functions (such as pedometer, tapping), etc.; as for other sensors that can be configured in the mobile phone, such as gyroscopes, barometers, hygrometers, thermometers, infrared sensors, etc., they will not be repeated here.

[0117] The audio circuit 660, the speaker 661, and the microphone 662 can provide an audio interface between the user and the mobile phone. The audio circuit 660 can transmit the received audio data to the speaker 661 after converting the received audio data into an electrical signal, which is converted into a sound signal for output; on the other hand, the microphone 662 converts the collected sound signal into an electrical signal, which is received by the audio circuit 660 and converted into audio data, and then the audio data is output to the processor 680 for processing, and then sent to another mobile phone through the RF circuit 610, or the audio data is output to the memory 620 for further processing.

[0118] WiFi is a short-range wireless transmission technology. The mobile phone can help users send and receive emails, browse web pages and access streaming media through the WiFi module 670. It provides users with wireless broadband Internet access. Figure 6 A WiFi module 670 is shown, but it is understandable that it is not an essential component of the mobile phone and can be omitted as needed without changing the essence of the invention.

[0119] The processor 680 is the control center of the mobile phone. It uses various interfaces and lines to connect various parts of the entire mobile phone. It executes various functions of the mobile phone and processes data by running or executing software programs and / or modules stored in the memory 620, and calling data stored in the memory 620. Optionally, the processor 680 may include one or more processing units; preferably, the processor 680 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 680.

[0120] The mobile phone also includes a power supply 690 (such as a battery) for supplying power to various components. Preferably, the power supply can be logically connected to the processor 680 through a power management system, so that the power management system can manage charging, discharging, power consumption and other functions.

[0121] Although not shown, the mobile phone may also include a camera, a Bluetooth module, etc., which will not be described in detail here.

[0122] In this embodiment, the processor 680 included in the terminal device also has the following functions:

[0123] Determining a training data set from historical media data according to historical user behavior data, wherein the historical user behavior data includes historical media data clicked by a user;

[0124] According to the training data set, a feature vector of the historical media data is obtained using a vector prediction model;

[0125] Counting the type identifiers included in the training data set, where the type identifiers are used to reflect the content characteristics of the historical media data;

[0126] Determining a feature vector of the type identifier according to a feature vector of historical media data corresponding to the type identifier;

[0127] If target media data for which a feature vector has not been calculated is obtained, a feature vector of the target media data is calculated according to a feature vector of a type identifier of the target media data.

[0128] This application embodiment also provides a server, see Figure 7 As shown, Figure 7 The structural diagram of the server 700 provided in the embodiment of the present application, the server 700 may have relatively large differences due to different configurations or performances, and may include one or more central processing units (CPU) 722 (for example, one or more processors) and a memory 732, and one or more storage media 730 (for example, one or more mass storage devices) storing application programs 742 or data 744. Among them, the memory 732 and the storage medium 730 can be temporary storage or permanent storage. The program stored in the storage medium 730 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the server. Furthermore, the central processing unit 722 can be configured to communicate with the storage medium 730 and execute a series of instruction operations in the storage medium 730 on the server 700.

[0129] The server 700 may also include one or more power supplies 726, one or more wired or wireless network interfaces 750, one or more input and output interfaces 758, and / or one or more operating systems 741, such as Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, etc.

[0130] The steps performed by the server in the above embodiment can be based on the Figure 7 The server structure shown.

[0131] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0132] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0133] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0134] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0135] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0136] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store program codes.

[0137] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A feature vector calculation method based on artificial intelligence, characterized in that: The method comprises: Determining a training data set from historical media data according to historical user behavior data, wherein the historical user behavior data includes historical media data clicked by a user; According to the training data set, a feature vector of the historical media data is obtained using a vector prediction model; Counting the type identifiers included in the training data set, where the type identifiers are used to reflect the content characteristics of the historical media data; Determining a feature vector of the type identifier according to a feature vector of historical media data corresponding to the type identifier; Target media data for which a feature vector has not been calculated is obtained, and a feature vector of the target media data is calculated according to a feature vector of a type identifier of the target media data.

2. The method according to claim 1, characterized in that If the target media data has historical user behavior data, after acquiring the target media data, the method further includes: According to the historical user behavior data of the target media data, a feature vector of the target media data is obtained through an incremental learning method.

3. The method according to claim 1, characterized in that If new user behavior data is generated for the target media data after the preset time, the method further includes: According to the new user behavior data, feature vectors of all media data are updated by an incremental learning method, where the all media data includes the target media data and the historical media data.

4. The method according to claim 1, characterized in that If the historical media data corresponding to the type identifier includes multiple pieces, determining the feature vector of the type identifier according to the feature vector of the historical media data corresponding to the type identifier includes: Acquire all historical media data corresponding to the type identifier from the training data set; The feature vectors of all historical media data are averaged as the feature vector of the type identifier.

5. The method according to any one of claims 1 to 4, characterized in that: If the historical media data has been viewed, the historical user behavior data also includes comment information.

6. The method according to claim 5, characterized in that Determining a training data set from historical media data according to historical user behavior data includes: Determining historical media data having positive comment information according to the comment information; The training data set is determined according to historical media data having positive comment information.

7. The method according to any one of claims 1 to 4, characterized in that: The obtained feature vectors of all media data are used to train a recommendation model, where the all media data include the target media data and the historical media data.

8. A feature vector calculation device, characterized in that: The device comprises a first determining unit, a second determining unit, a statistical unit, a third determining unit and a calculating unit: The first determining unit is used to determine a training data set from historical media data according to historical user behavior data, wherein the historical user behavior data includes historical media data clicked by a user; The second determining unit is configured to obtain a feature vector of the historical media data using a vector prediction model according to the training data set; The statistical unit is used to count the type identifiers included in the training data set, and the type identifiers are used to reflect the content characteristics of the historical media data; The third determining unit is used to determine the feature vector of the type identifier according to the feature vector of the historical media data corresponding to the type identifier; The calculation unit is used to obtain target media data for which feature vectors have not been calculated, and calculate the feature vectors of the target media data according to the feature vectors of the type identifications of the target media data.

9. The device according to claim 8, characterized in that If the target media data has historical user behavior data, after acquiring the target media data, the device further includes: a fourth determining unit; The fourth determining unit is configured to obtain a feature vector of the target media data by an incremental learning method according to historical user behavior data of the target media data.

10. The device according to claim 8, characterized in that If new user behavior data is generated for the target media data after a preset time, the device further comprises: an updating unit; The updating unit is used to update the feature vectors of all media data by an incremental learning method according to the new user behavior data, wherein the all media data includes the target media data and the historical media data.

11. The device according to claim 8, characterized in that If the historical media data corresponding to the type identifier includes multiple pieces, the third determining unit is specifically configured to: Acquire all historical media data corresponding to the type identifier from the training data set; The feature vectors of all historical media data are averaged as the feature vector of the type identifier.

12. The device according to any one of claims 8 to 11, characterized in that: If the historical media data has been viewed, the historical user behavior data also includes comment information.

13. The device according to claim 12, characterized in that The first determining unit is configured to: Determining historical media data having positive comment information according to the comment information; The training data set is determined according to historical media data having positive comment information.

14. The device according to any one of claims 8 to 11, characterized in that: The obtained feature vectors of all media data are used to train a recommendation model, where the all media data include the target media data and the historical media data.

15. A device for calculating a feature vector, characterized in that: The device comprises a processor and a memory: The memory is used to store program code and transmit the program code to the processor; The processor is configured to execute the method according to any one of claims 1 to 7 according to the instructions in the program code.

16. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store program codes, and the program codes are used to be executed by a processor to implement the method according to any one of claims 1 to 7.

Images